This cross-sectional study was conducted using data from the 2015-2016 MDHS. The survey is part of the global Demographic and Health Survey (DHS) program, which assists countries in periodically collecting data to monitor and evaluate populations and health and nutrition programs. The MDHS 2015-2016 was implemented by the National Statistical Office from October 19, 2015 to February 18, 2016. The survey involved interviews with a representative sample of women and men, and collected information on households, data on biomarkers, and anthropometric measurements of women (15-49 years of age) and children under 5 years of age, among other information. As part of this survey, 24,562 women aged 15-49 years were interviewed. We excluded 1,883 women who self-reported that they were pregnant, 3,022 who were below 18 years of age and 13,254 who had missing or flagged height and weight measurements, or for whom anthropometry was not done for other reasons (e.g., refusal). Pregnant women were excluded because pregnancy nullifies results for body mass index (BMI) as an indicator of nutrition status, especially at the population level, and women aged 15-17 years were excluded because they are still undergoing rapid physical changes, meaning that BMI alone is not a good indicator of their nutritional status (in this case overweight and obesity). The result was a final analytical sample of 6,443 adult non-pregnant women aged 18-49 years. Our study involved an exclusive population of women only, because anthropometric data for the MDHS was collected only in women. In addition, globally, more women than men are overweight or obese, and maternal overweight and obesity are associated with numerous health consequences and an emerging cycle of these conditions across multiple generations if left unchecked [11].

Height and weight were recoded following the Centers for Disease Control anthropometry protocol [12]. Height was measured in a standing position using standard height boards, while the participant was in a relaxed state. Shoulders were stress-free, and hair pins and all other head buns and ornaments were removed. Each woman was directed to stand straight, with her back against the board, toes apart, and heels together. Height was measured to the nearest 0.1 cm. A calibrated digital scale was used to measure weight. The scale was put on a flat surface and tared to zero. The participant was requested to put on light clothing, and stand straight on the scale, with eyes focused in front. Weight was measured to the nearest 0.1 kg. Height and weight were used to estimate the prevalence of overweight and obesity using BMI (calculated as a continuous variable using the following formula: weight [kg]/height [m]²).

The main outcome variable was overweight and obesity (BMI ≥25.0 kg/m²). For the regression analysis, BMI was recoded into a binary variable with the categories of not overweight or obese (BMI ≤24.9 kg/m²) and overweight or obese (≥25.0 kg/m²). For other ancillary analyses, BMI was categorized based on the WHO [13] classes as follows: underweight (<18.5 kg/m²), normal (18.5-24.9 kg/m²), overweight (25.0-29.9 kg/m²), obese (≥30.0 kg/m²), pre-obese (25.0-29.9 kg/m²), obese class I (30.0-34.9 kg/m²), obese class II (35.0-39.9 kg/m²), and obese class III (≥40.0 kg/m²).

The covariates (social and demographic associates) were presented and defined as follows: age in years (calculated as a continuous variable and stratified into 6 categories: 18-22, 23-27, 28-32, 33-37, 38-42 and ≥43 years), area of residence (rural or urban), marital status (recoded into 6 categories: never in union, married, living with partner, widowed, divorced, and separated), educational status (categorized as no education, primary education, secondary education, and higher education), occupation (recoded and categorized into 8 categories based on the International Standard Classification of Occupations [14]: not working, professional/managerial/technician, clerical/sales, agriculture, household and domestic, services, skilled manual and unskilled manual), socioeconomic status (SES) (indicators of SES were assessed using the wealth index quintile, ranging from first to fifth, which translate to poorest to richest in that order), health insurance (yes or no regardless of the insurance source), parity (number of births in the previous 5 years, categorized as none, 1, or ≥2 births), total number of people living in the household (3 or fewer, 4-7, 8-10, or ≥11 people), current smoking status (does not smoke or currently smokes), contraceptive use (using any form of contraceptive or not using any form of contraceptive, religion (recoded into 5 categories: none, other Christian, Muslim, Catholic, and Protestant), Ethnicity (recoded into 6 categories: others, Tumbuka, Lhomwe, Yao, Ngoni, and Chewa).

The analysis was carried out in SPSS version 20.0 (IBM Corp., Armonk, NY, USA) for descriptive statistics (frequencies and weighted percentages) and binary logistic regression to evaluate associations and determine odds. Frequency was analyzed and presented as unweighted, but percentages were weighted so that the results represent the Malawian adult non-pregnant women population. The prevalence estimates for underweight, normal weight, overweight, and obesity were computed as percentages with the total sample size as the denominator. Furthermore, the prevalence of overweight and obesity and their associations with different categorical background characteristics were analyzed using the chi-square test. Results were considered statistically significant at p-values <0.05. Binary logistic regression was used to determine possible relationships and the direction of the associations between different independent variables and the outcome. Bivariate binary logistic regression yielded crude odds ratios (with 95% confidence intervals [CI]) for membership in the overweight or obese category. This was followed by multivariate logistic analysis to adjust for all the social and demographic determinants of the study (independent variables), thereby generating adjusted odds ratios (aOR) and 95% CIs.

Prior to conducting this research, a request for approval to download and use datasets (MDHS 2015-16) was sort from the DHS programme. The programme’s institutional review board granted authorization (AuthLetter_120130). On the part of the data source, DHS maintains good ethical standards in all the surveys it conducts. The MDHS 2015-16 was preceded by a national campaign to raise public awareness of the survey aims and give information about its process. Participation in the survey conducted was completely voluntary and with full autonomy to take part or to reject participation at any point of the survey. No perceived potential harm was encountered on refusing participation. Informed, verbal individual consent was obtained from respondents before conducting the questionnaire or interview. All participants’ information were processed anonymously and labeled with just Identification codes. The survey was supervised and approved by the Ministry of Health.

The background characteristics of the study subjects are presented in Table 1. Of all the participants, 81.5% were from rural areas, 68.0% were married, and 59.0% had completed up to primary-level education. The 18-22 age group accounted for 23.7% of the participants, while the lowest percentage was found in the 38-42 age group (10.9%). With regard to religion, 26.4% of the respondents were Protestants, 17.5% Catholics, 12.6% Muslims, and over a quarter (42.9%) belonged to other forms of Christianity. Furthermore, 23.6% of the subjects were in the fifth wealth index quintile, indicating the highest level of SES, 19.3% were in the poorest (first), 19.2% were in the fourth quintile (rich), 19.1% were poor (second quintile), and 18.8% were in the middle class (third quintile).

Table 2 presents a summary of the anthropometric measurements of study subjects. The subjects in rural areas were marginally older than those in urban areas (30.00±8.70 years vs. 29.00± 7.80 years, respectively). The mean weight, height, and BMI of women were 56.31±10.40 kg, 1.56±0.06 m, and 23.06±3.94 kg/m², respectively. Differences in background characteristics were significantly associated with BMI, except for region of residence (p<0.001 vs. p=0.056) (Table 3). Overall, 5.7% of the women fell into the <18.5 kg/m² (underweight) category of BMI, with no significant differences by area of residence (5.2% in urban areas vs. 5.8% in rural areas; p=0.046; comparison with an alpha of 0.006) (Table 3). Approximately 16.8% of the respondents fell into the BMI category of 25.0-29.9 kg/m² (pre-obese/overweight). The distribution of overweight was significantly different between urban and rural residents (25.5% in urban areas and 14.8% in rural areas; p<0.001). Overweight was almost equally pronounced in women aged 28-32 years and those aged 33-37 years (21.6% and 21.4%, respectively). Over a quarter (26.1%) of women with a higher level of education and those in the richest wealth quintile (25.8%) were overweight. Overall, 6.3% of women were in the obese (BMI ≥30.0 kg/m²) category, with significant differences between rural and urban areas (14.6% in urban areas vs. 4.4% in rural areas; p<0.001). Obesity was more pronounced in women aged 33-37 years (11.6%), those with a higher level of education (15.2%), those who had a professional/managerial/technician occupation (15.9%), and those in the richest SES group (14.6%).

Overall, the prevalence of overweight and obesity among adult women in the study was 16.8% and 6.3%, respectively. In urban areas, 25.5% and 14.6% of the women were overweight and obese, respectively. In contrast, in rural areas, 14.8% and 4.4% of the women were overweight and obese, respectively.

An ancillary categorization of obesity (BMI ≥30.0 kg/m²) (Figure 1) by area of residence showed that 4.9% of the respondents were obese class I, 1.0% were obese class II, and 0.3% were obese class III. In urban areas, 11.0%, 2.5%, and 1.2% of the respondents were obese class I, obese class II, and obese class III, respectively. In rural areas, 3.6% of the respondents were obese class I, 0.7% were obese class II, and 0.1% were obese class III. The distribution of pre-obese, obese class I, obese class II, and obese class III women was also considered in reference to SES strata (Table 4). Approximately 11.3% of women in the highest SES stratum were obese class I, and more than a quarter (25.8%) were pre-obese. In the poorest stratum, the prevalence of pre-obese women was 10.9%, marginally higher than in poor women (10.8%).

Table 5 shows the results of binary logistic regression for the predictive probability for membership in the category of overweight or obese. Women who were in the age group of 33-37 years were 4 times more likely to be overweight or obese than those aged between 18 and 22 years (aOR, 3.95; 95% CI, 2.91 to 5.36; p<0.001). The odds of a woman residing in urban area being overweight or obese were 18.0% higher than those of a woman residing in rural area (aOR, 1.18; 95% CI, 0.93 to 1.49). Married women were less likely to be overweight or obese than those who had never been married. The crude odds ratios suggested that the odds of being overweight or obese increased with increasing levels of education, but this was not confirmed in the multivariate analysis. Women from the Ngoni ethnic group were more likely to be overweight or obese than women of other ethnic groups in Malawi (aOR, 1.54; 95% CI, 1.14 to 2.08; p<0.05).

Women in agricultural occupations (self-employed or employee) were less likely to be overweight or obese, which was highly statistically significant (aOR, 0.69; 95% CI, 0.57 to 0.84; p<0.001), while women who were skilled manual workers were more likely to be overweight or obese than those who were not working (aOR, 1.59; 95% CI, 0.96 to 2.64). The risk of being overweight or obese increased with increasing SES, as women belonging to the richest (fifth quintile of the wealth index) group were over 3 times more likely to be overweight or obese than those in the poorest quintile (aOR, 3.30; 95% CI, 2.46 to 4.43). Coverage by any form of health insurance was associated with being overweight or obese, and current smokers were less likely to be overweight or obese. A higher likelihood of being overweight or obese was seen in women residing in households with 8-10 people, as compared to those having 3 or fewer people (aOR, 1.09; 95% CI, 0.81 to 1.46).